How to make a winning robot
For our final robotics project at Swinburne we decided to compete in the first annual Autonomous Robotics Competition hosted by National Instruments. The competition involved designing, building and programming an autonomous robot to perform tasks that would be required in an autonomous warehouse system.
The actual task was to navigate a 4x4 meter arena while avoiding obstacles and locate three ‘packages’ which needed to be transported to their respective drop off zones. This task had to be completed in less than 10 minutes with points being deducted for any mistake – such as hitting an obstacle or dropping a package in the wrong zone.
The robot had to be controlled using a CompactRIO, this is an industry quality controller that has both a real time target and field programmable gate array that can communicate effortlessly, giving us high processing capability and ease of implementation. The CompactRIO had to be programmed with LabView, a data flow programming language, which we had to learn as none of the team members had any experience with this technology. We quickly discovered that, although different ton anything we had done before - LabView allowed programming in a very intuitive manner and freed us to concentrate on the higher lever logical problems rather than worrying about bits and threads.
Mechanically we think our robot is brilliant. We opted to pick up all the packages at once and store them so that we could save time. The design to accomplish this took quite a lot of brainstorming to develop. We wanted it to be as robust as possible and simple to manufacture and control. The final design included a very simple two degrees of freedom gripper that would pick up the packages and place them on a ramp, with gates that would open and close to control the packages sliding off the back.
The chassis was made from 3mm plate aluminium and was based on a traditional two wheel differential steering frame, allowing for nimble, precise and predictable movements.
We decided to keep the sensing as simple as possible, with the final robot only having three infrared range finders and two reflectance sensors. Cameras and light detection and ranging technology could also have been used but we felt they were unnecessary and would complicate the build too much.
As for the control of the robot, we used Adaptive D* path planning to map the locations of the obstacles in the arena and plan the most efficient paths around them. The robot was steered using the Pure Pursuit algorithm along these generated paths. The position tracking was done with a combination of Dead Reckoning, and position/heading snapping as the robot would drive over the tiled arena.
We attacked the problem as soon as the competition details were released. Many prototypes were built to test sensors, logic and the mechanics of the design. And once the final robot was built we had two months to test everything thoroughly - we would not have wanted any less time as we were still discovering issues just days before the competition.
The competition was held in Sydney on Friday 16 September, and we won. Ten universities from across Australia signed up for the competition, with only six of them attending the event. National Instruments organised a few guest speakers to talk about where they saw the future of Robotics, and lunch was provided. Our team competed third and our robot performed perfectly - completing the competition in just over three minutes without making a single error. RMIT came second with a time just behind ours and a few minor mistakes. The day was great, but nerve wracking, as so many influential people were watching.
This has been a great experience for us and has provided great opportunities in the future, on top of the $3000 prize. National Instruments has also given us each a copy of LabView, the option to get LabView certification for free, and they have asked us to speak at their annual conference next year. We would like to thank Swinburne for financially supporting us and for giving us the room to set up an arena, Mark Phillips and Mehul Satikunvar for organising the competition and Maxon Motors for providing the motors.
Ben Smith, Jeremy Wu and Nic Dortmans
View the Swinburne media release.
3 notes
swinburne posted this